Miller 2020.

Study characteristics
Methods	Study design: cluster‐RCT Study grouping: parallel group Cluster size: 2 (parent level) Study duration: 10 weeks
Participants	Inclusion criteria: Syrian refugee or vulnerable host community families with ≥ 1 child aged 3–12 years; both primary caregivers willing to participate in the study and willing to commit to attending all 9 sessions of the Caregiver Skills Intervention if randomised to the Caregiver Skills Intervention arm of the study; participating caregivers were Arabic speaking Exclusion criteria: prior or current participation by either caregiver in a parenting or stress management intervention; family did not have a child aged 3–12 years; anyone who was unable, even with assistance, to complete the assessment questionnaires; unwillingness of either caregiver to give informed consent Age range: children 3–12 years; parents not specified Gender (parents): 52% women, 48% men Gender (children): 41% girls, 59% boys Intervention sample: 78 Control sample: 73 Main type of traumatic event: displacement Phase of humanitarian crisis: during the acute crisis (mortality was still higher than it was before the crisis). The study took place between November 2017 and July 2019. Type of humanitarian crisis: war/armed conflict
Interventions	Intervention name: Caregiver Support Intervention (CSI) Delivered by: para‐professionals: non‐mental health specialist, aged ≥ 24 years, with ≥ 2 years of experience implementing psychosocial interventions, preferably with adults, even more preferably with parents/caregivers Format of therapy: face to face Number of sessions (total): 9 Type of control: waiting list Type of intervention context: group intervention – offices of 3 Community‐Based Organizations (CBOs) with which War Child Holland (WCH) collaborates in the target communities Type of promotion intervention: group level Description of the intervention: 9‐session, weekly group intervention, offered separately to women and men. Sessions 1–4 focussed exclusively on caregiver well‐being (covering topics such as stress and relaxation, lowering stress, and coping with frustration and anger). Sessions 5–8 focussed on strengthening parenting under conditions of adversity, and draw heavily on social learning theory and commonly used methods of training in positive parenting (i.e. increasing awareness of the impact of stress on parenting, increasing positive parent‐child interactions and the use of non‐violent discipline methods, and reducing harsh parenting). Session 9 entailed a review and closing of the intervention. In addition, in each session, participants were introduced to a new relaxation or stress management technique. These techniques were also provided to participants in Arabic on mp3 files, which they could either play on their smartphones or on mp3 players provided at the start of the programme. Participants were encouraged to practice any relaxation or stress management activity ≥ 3 times each week.
Outcomes	The cluster effects were controlled by using the STATA command "clustersampsi." Mental well‐being Outcome type: continuous outcome Reporting: fully reported Scale: Kid‐KINDL for Parents Direction: higher is better Data value: endpoint Acceptability (dropout from trial) Outcome type: dichotomous Reporting: fully reported Data value: endpoint
Notes	Sponsorship source: funding grants from the Bernard van Leer Foundation and Open Society Foundations. Ethical approval provided by the University of Balamand, Tripoli, Lebanon Country: Lebanon Setting: city of Tripoli in North Lebanon. 70,000 registered Syrian refugees were living in Tripoli Author's name: Kenneth E Miller Institution: War Child Holland Email: kenneth.miller@warchild.nl Address: Helmholtzstraat 61g, 1098LE Amsterdam, The Netherlands
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "After participants had completed the baseline assessment, our research coordinator explained the randomization process to participants, inviting them to draw a lollipop out of an opaque bag containing an equal number of red and green lollipops, corresponding to the number of participants at the assessment. Baseline assessment took place over five days at the three CBOs, so this process was repeated several times. The first caregiver to be assessed from each family drew the lollipop that determined that family's group assignment. Once all data were collected from the full sample, the group assignment represented by each color was determined by a coin toss done by a WCH staff member based in Amsterdam who was unaffiliated with the study."
Allocation concealment (selection bias)	Low risk	Group assignment was concealed until a coin was flipped in Amsterdam.
Blinding of participants and personnel (performance bias) All outcomes	High risk	The Lebanon‐based research co‐ordinator was not blind, neither, by necessity, were participants blind to their group assignment.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The principal investigator, co‐investigators, and research assistants remained blind throughout the study.
Incomplete outcome data (attrition bias) All outcomes	Low risk	2/151 participants dropped out from the study.
Selective reporting (reporting bias)	Low risk	All measures described in the methods section of the article were also reported in the results. No trial protocol available, but trial registered at the ISRCTN registry (ISRCTN33665023).
Therapist qualification	Low risk	Quote: "non‐mental health specialist, 24 years or older, with at least 2 years of experience implementing psychosocial interventions, preferably with adults, even more preferably with parents/caregivers, emotionally mature."
Therapist/investigator allegiance	Unclear risk	No information provided.
Intervention fidelity	Low risk	Quote: "A review of the checklists indicated that all activities were implemented as designed in six of the seven groups. In one of the men's groups, however, the facilitators were initially insufficiently prepared and failed to implement two activities in each of the first two sessions."
Other bias	Low risk	No other sources of bias detected. Cluster‐RCT risk of bias extension 1. Recruitment bias; the recruited population belonged to the same catchment area (low risk of bias) 2. Baseline imbalance; cluster balance was maintained after randomisation (low risk of bias) 3. Loss of clusters; just 1/79 clusters were lost (low risk of bias) 4. Incorrect analysis; the analyses were correctly conducted and reported (low risk of bias)